攻城狮/业余投机/右侧交易/游戏开发/Haskell/Rust/C++/Unity3D/C#/对java有偏见/RL/智商欠费/浅尝辄止故平庸/反乌托邦/竹林中/地狱变/手撕菠萝蜜/胸口碎榴莲/单机推特中/乐视一生黑（乐视已阵亡）/华为一生黑（迟早会阵亡 ）/中华跪族/器材党/预防式B台支
上海, 中华人民共和国
Based in United States
- Tweets 5,242
- Following 2,489
- Followers 1,240
- Likes 33,640
See? What did I say?
Value for money and word of mouth are the best marketing. Kimi’s model isn’t bad, but its usage quota is the lowest among all coding plans I’ve subscribed to.
That's how Kimi AI are promoting their new model Kimi K3.1 pro and I know all the influencers working with them and none of them are honest,I have posted about it myself but I didn't join this and never used their templates that they provide.
PROPAGANDA, I have more to share it's f**cked up I'm gonna have follow up for this
2
13
1,994
WTF！I generated a model of DeepSeek chan ( 大肥鱼 | 鲸鱼娘) using kimi 3.1 pro in a short prompt. 3D artists will be cooked.
Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.
🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.
Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API.
Open weights by Nov 4, 2026.
🔗 API: platform.kimi.ai
🔗 Tech blog: kimi.ai/blog/kimi-k3-1-pro
3
18
7,663
Kimi 3D Make is crazy. It generated DeepSeek chan (大肥鱼 | 鲸鱼娘) given a picture.
It outputs printable model. You can view it online :
2r6dhnhmb6ani.ok.kimi.link/
9
1
60
20,635
A few months ago, I heard that Kimi's next-generation model would be named with "Pro." I thought it would be K3 Pro. But judging from the blog page that appeared and then disappeared today, it's called K3.1 Pro. 😂
12
1,223
Take a look at what DeepSeek-V4.1-Flash-1017 also dug out of Kimi’s js: BOT.
Yes, Kimi will probably also launch a bot aligned with Grok Bot and Dots—so what will happen to the original KimiClaw? What differences will the new bot mode have?
Kimi has seen how passionate everyone is about creating 3D games and video showcases, so the new version adds a Make 3D feature—and we’re really looking forward to it.
20
1,822
Kimi has seen how passionate everyone is about creating 3D games and video showcases, so the new version adds a Make 3D feature—and we’re really looking forward to it.
We can see `k3d1` (Kimi K3.1) model and `make3d` mode of Kimi's "OK Computer" in the model selection js now. Get hyped!
statics.moonshot.cn/kimi-web…
36
3,973
An intriguing paper reveals periodic positional blind spots in DeepSeek-V4/V4.1 caused by chunked KV-cache compression, where shifting content by just a few tokens can dramatically reduce retrieval accuracy—a problem significantly mitigated, but not eliminated, in V4.1.
paper: arxiv.org/abs/2609.36322
interactive demo: ultimatejupiter.github.io/bl…
9
543
StepFun is cooking
Step 5 Preview now exists where you build.
Now live on @OpenRouter. A week of free access is rolling out today across @opencode, @cline , @NousResearch, @kilocode, and more.
Flagship-tier intelligence for agentic and professional work, at substantially lower task cost.
Keep your workflow. Switch your model.
Try Step 5 Preview on OpenRouter: openrouter.ai/stepfun/step-5…
1
1
8
2,709
Gemini 4 returned to third place after more than two months of training. It must be said that the gap between China and the U.S. in compute scale is widening.
Gemini 4 is already cooking—pretraining’s underway. The biggest twist this time? They’re really sweating the running costs now. Gemini used to play the bang-for-your-buck card, which kinda left it in the dust in the race for state-of-the-art.
5
375
CXMT (688825.SH) announced plans to invest RMB 24.1bn in a new technology R&D project, including RMB 13bn in excess IPO proceeds.
It also plans to invest RMB 10.8bn in Phase II of its memory wafer back-end testing base, including RMB 5bn in excess IPO proceeds. The new facility will provide DRAM chip testing and module assembly.
Total planned investment: RMB 34.9bn (~US$4.9bn).
#CXMT #DRAM
2
189
The previously missing data from MiMo’s RL training has now been added.
MIMO’s RL training has stopped at step 30. We can observe the following:
The Pro model achieved a DeepSWE score of 72.57, but this score was reported at step 28; data for steps 29 and 30 are missing.
The number of active environments for Pro began increasing at step 23, then dropped rapidly after step 26. The duration of each training step also rose sharply.
5
647
MIMO’s RL training has stopped at step 30. We can observe the following:
The Pro model achieved a DeepSWE score of 72.57, but this score was reported at step 28; data for steps 29 and 30 are missing.
The number of active environments for Pro began increasing at step 23, then dropped rapidly after step 26. The duration of each training step also rose sharply.
2
5
1,036
Kimi K2.8 Preview offers a 1M context window, and its performance is officially claimed to be close to K3.
kimi.com/code/docs/en/kimi-c…
11
1,068
DeepSeek-V4.1-Flash delivers beastly performance at a fraction of Kimi K3’s price. ↓
DeepSeek-V4.1-Flash: 552B parameters, with just 8B active during prefill and 16B during decoding. Kimi K3: 2.8T total, 104B active. Those benchmark numbers make V4.1-Flash look like an absolute beast!
And judging by the release cadence, Kimi’s next model—whether it’s K3 Pro or K3.1—should be just around the corner.
5
304
DeepSeek-V4.1-Flash: 552B parameters, with just 8B active during prefill and 16B during decoding. Kimi K3: 2.8T total, 104B active. Those benchmark numbers make V4.1-Flash look like an absolute beast!
And judging by the release cadence, Kimi’s next model—whether it’s K3 Pro or K3.1—should be just around the corner.
1
12
962
DS has a very interesting question in the V4.1 Flash survey:
"Do you think this model can fully replace the online (production) version of DeepSeek V4 Pro?"
It seems that high expectations are being placed on this multimodal V4.1 Flash.
"DeepSeek V4.1 Flash is now in internal beta as an intermediate version — welcome to try it out. It features a new model architecture with native multimodal support, stronger capabilities, faster speed, and lower cost.
Keep your base_url unchanged and set the model name to deepseek-v4.1-flash-expires-on-0910 to start using it. Pricing is currently the same as deepseek-v4-flash, with a rate limit of 20 concurrent requests per account."
2
33
38,798
Xiaohongshu (RedNote) just dropped its multimodal language model. It has text, image, video, and audio understanding, with 280B / 16B active parameters and a 512K context window. What surprised me is that it seems to outperform MiniMax M3 (428B / 22B active). 👇
Introducing dots3-note preview — a small but mighty step toward long-horizon agency in real life.
🔹 280B MoE with 16B active parameters, a 512K context window, and multimodal understanding across text, vision, and audio
🔹 Introduces TEMPO, a new RL approach for long-horizon agent training through self-critiquing and test-time-scaled value estimation
🔹 Built to reason, explore unfamiliar environments, update memory over time, and combine multimodal perception with coding and tool use to solve complex tasks
🔹 Open weights on Hugging Face, alongside two open benchmarks for real-life agents: VibeSearchBench and VibeLifeBench
Competitive with much larger models across reasoning, agentic, and multimodal evaluations.
🔗 Tech blog: studio.dots.ai/dots/dots3-en…
🔗 Model weights: huggingface.co/dots-studio/d…
🔗 Github: github.com/studio-dots-ai/do…
1
326
GLM-5.3 is finally out. Big upgrade from 5.2, but clearly the market was expecting even more. Oh well, bring on GLM-5.5.
I heard that 5.5 would be a major version update, but in the management conference call the other day, there was also mention of a minor update, 5.3. Now it looks like 5.3 might be abandoned.
3
39
5,864
GLM-5.3
Post-training keeps working.
Introducing GLM-5.3: Built to Code. Ready for Cyber Defense.
- Top-tier coding and agentic capabilities, achieved through post-training on the 743B base model
- A major leap in cybersecurity, setting a new standard among open models
Tech Blog: z.ai/blog/glm-5.3
2
2
20
2,852
